I. Introduction

Millions of head injuries occur each year. The World Bank estimates that approximately five million head injuries per year can be attributed just to traffic accidents worldwide. A sizeable fraction of the patients die, and a sizeable fraction survive with severe, long-term disabilities. Recognizing the need to optimize the treatment and care of patients with traumatic head injury, NINDS sponsored a workshop of experts from academia and industry. The goals of the workshop were to review results from recent clinical trials, to apply lessons learned from these trials in designing new trials, and to spur continued research and development efforts from both academic scientists and the pharmaceutical industry.

Report of Working Group on Clinical Trials in Head Injury

II. Background

Traumatic brain injury (TBI) is a complex disease, with a variety of precipitating causes that affect the location and severity of injury and influence the course of recovery. Tissue damage associated with head injury include axonal injury, focal contusions and edema, and intracranial hematomas and swelling. To repair the original injury is acknowledged to be difficult, but the spread of secondary damage to the brain can possibly be contained.

Many physiological processes have been implicated in contributing to brain damage. While some axonal damage occurs immediately upon injury, delayed axotomy has been correlated with the activity within damaged cells of calcium-activated proteases called calpains. A major contributing factor for such secondary damage to axons and cells is thought to be cerebral ischemia. Low cerebral blood flow, low oxygen tension, and release of excitatory amino acids are all associated with ischemia and can trigger injury cascades within cells. Edema, originally thought to be due to leakage from blood vessels and breakdown of the blood-brain barrier, is now considered to be a consequence of cell death in the injured tissue. The interactions among these various physiological events, and the long-term changes over time, are still unclear.

Several prognostic factors are thought to be reliable in predicting outcome based on a patient's acute presentation and ICU course. These include age, pupillary reflexes, motor response, length of coma, intracranial pressure, and data from computed tomography (CAT scan). Consideration of such information is vital for effective design and interpretation of clinical trials.

III. Workshop Discussion

Results of recent clinical trials in TBI were reviewed by the participants, and several common themes and problems were highlighted. Among the most prevalent concerns were: What are the most relevant mechanisms of injury/damage to test? What are the best ways to translate results from animal models to clinical testing? What are the appropriate patient populations in which to test these mechanisms? What the right outcome measures to use?

Phase II/III trials of pharmacologic agents covered a wide range of target mechanisms, including glutamate excitotoxicity (selfotel, cerestat, CP101,606, EAA 494, dexanabinol), calcium-mediated damage (nimodipine, SNX-11), and lipid peroxidation (tirilazad, PEG-SOD). Moderate hypothermia was investigated in Phase II and Phase III trials. Smaller clinical studies or early Phase II trials also investigated the influence of factors such as IGF-1/growth hormone, anticonvulsants, bradykinin, and cerebral perfusion pressure.

In the majority of trials, the Glasgow Outcome Scale (GOS)¹ was used to measure patient outcome, and effectiveness of treatment was set at 10% increase in favorable outcome. For a variety of reasons, no clinical trial in head injury has been effective at this level. Several trials were terminated early, due to mortality in drug-treated groups, to interim analyses indicating no efficacy, or to disruption of study drug supply. In other trials, the treatment may have improved a measure of the targeted mechanism, but did not change poor neurological outcome. For example, treatment with IGF-1/growth hormone improved nitrogen retention; maintenance of cerebral blood flow reduced the number of ischemic episodes, and hypothermia decreased intracranial pressure, but in none of these trials was GOS improved in treatment versus control groups. A recurrent theme in the discussion of the various trials was that the preset goal of 10% increase in favorable outcome on the GOS was not sensitive enough, and that a need for other outcome measures exists in the field.

The heterogeneity of the patient population was recognized as a potential problem for interpreting results, and several trial designs stratified patients according to CT parameters or did not include the most severely injured patients. An imbalance in gender was also noted, hindering analyses for women.

Practical issues of conducting the trials were also addressed by the presenters. Reflecting the concerns of the preceding scientific review, outcome measures received much attention. Insight and some caution came from the FDA about the careful use of surrogate endpoints or physiological outcomes. Another important consideration concerned the theoretical therapeutic windows for targeted mechanisms, and the speed with which patients can be transported, worked up in the Emergency Department, and informed consent obtained from next of kin.

It was noted that while pharmacological intervention is focused at the cellular level, where drugs act on physiological mechanisms, patients and care-givers are most about disability and handicap. These outcomes operate at higher functional and social levels. Recovery continues over the long-term; however, current concepts of "windows of opportunity" for therapy target acute events. Expanding efforts to encompass acute, sub-acute, and chronic periods after TBI may open new avenues for clinical trials and basic research.

Finally, the management of head injuries in the intensive care unit (ICU) has a large impact on outcome, and must be considered when assessing any experimental intervention. The American Association of Neurological Surgeons (AANS) established guidelines for the acute and sub-acute treatment of head injuries in 1995². A recent survey of trauma centers in the United States by the Brain Trauma Foundation, found that approximately 60% of centers complied with the guidelines and these centers reported better outcomes as compared to other centers. An interactive internet database, which could serve as a source for data, on-line education for practioners, and centralized interpretation of CAT scans, was suggested as a possible approach to this issue.

IV. Recommendations/Conclusions

The goal of the workshop was to design better clinical trials, using lessons learned from previous trials. Recommendations for an optimized trial design, criteria for pharmaceutical sponsors, and proposals for two new trials are outlined below.

Optimal trial design features

1. Tailor protocol to a mechanism relevant to human head injury: Further basic research is needed to clarify the key mechanisms underlying mild, moderate, and severe head injuries. In each broad category, injury may be focal or diffuse, and changes are seen in morphology, blood flow, metabolism, receptor expression, and behavior. There are many different ways to produce head injury in experimental animals. No one animal model will provide the perfect correlation to human head injury; however, by consensus among investigators a few models can be chosen and standardized, and outcome data from them shared and evaluated.

   Basic research in animal models should emphasize:

   1. developing models for coma
   2. evaluating gender and age differences in response to head injury
   3. establishing animal ICUs
   4. developing models for acute therapies to be used in the field and emergency room
   5. addressing later events and new treatment windows
   6. developing models of secondary insults
   7. evaluating interactions with ancillary treatments (i.e., anticonvulsants)
   8. measuring pharmacodynamic and pharmacokinetic relationships
   9. investigating neurorestorative/plasticity mechanisms

   Clinical trials should use protocols that explicitly test the mechanisms relevant to patients with head injuries. Care should be taken to translate results "from bench to bedside" as closely as possible. This effort may require the use of acute therapies very early after injury--perhaps before arrival at the Emergency Room. Depending on the targeted mechanism, development of physiological outcomes or validation of surrogate endpoints may be necessary for future trials.

2. Obtain adequate preclinical data: Before carrying out clinical trials, sufficient preclinical data should be gathered for new pharmaceutical agents. This data should include the demonstration of:
   1. efficacy in at least 2 rodent models, and in larger animals (e.g., pig, dog, monkey) if possible. Efficacy results should ideally document that the compound in question works on a mechanism present in human TBI.
   2. transport across the blood-brain barrier and adequate concentrations for activity in the brain
   3. behavioral, physiological, and histological data to support further development of the drug, including the long-term effects of the compound and its activity in mild, moderate, and severe injuries

3. Focus the trial on an appropriate subpopulation of patients: Heterogeneity within the head trauma patient group is broad. To compare active and placebo groups, inclusion criteria should more specifically state requirements for factors such as severity of injury and prognostic variables, or patients should be stratified into different groups for analysis. Targeting the right patient subgroup would increase the power to detect treatment differences, and would allow smaller sample sizes. Some pathological processes may only be relevant a sub-population of patients and not to the entire diverse group of patients. Phase II trials need to be conducted to identify the subgroup that matches the mechanism.

4. Choose the right outcome measures: Prerequisites for a useful outcome measure are that it must be ordinal, easy to interpret, assessed in all patients, and the misclassification error must be small. It is important to obtain a complete set of data, so care should be taken to ask for the right amount of data, and to make sure that the data is collected from every patient.

   The 5-point Glasgow Outcome Score (GOS) is widely used as a primary outcome in clinical trials. The GOS has recently been expanded to include further divisions in the disability categories, and use of the expanded GOS at the 6-month time point is recommended. In practice, the GOS is frequently divided into "favorable" outcome (MD and GR) and 'unfavorable' outcome (D, V, SD). Detecting a treatment difference between these two large groupings has proven to be difficult. Refining the expected outcome to suit the population being studied, for example by breaking the dichotomy at a different place (GR vs. MD, SD, V and D), or looking for shifts from one category to another, may be a more useful way to look for treatment effects. Another approach is to relate outcome to prognosis. Available data should be re-analyzed so that strong prognostic factors can be identified, and if their predictive value is accepted, it would be possible to compare expected versus observed outcome in response to the therapy under investigation.

   In many clinical trials of head injury, demonstration of effectiveness was set at 10% increase in the percentage of patients with favorable outcome on the GOS. This level has not been achieved, although trends towards favorable outcome have been observed in some trials. The 10% bar may be unrealistic: approximately 30% of patients achieve good recovery regardless of treatment. Should trials look for smaller (5% or 7.5%) increases in favorable outcome? Effect sizes of only 3-5%, would translate into 1000-5000 fewer deaths from TBI each year. Setting the bar lower may achieve more "positive" outcomes of clinical trials; however, the approval of mediocre therapies may hinder the development of truly effective treatments. Research must be directed at exceeding standard therapy. From a regulatory point of view, the FDA is not concerned about the size of the treatment effect, and a proposed intervention must simply be clearly different from control.. Treatment effect size is a question of clinical usefulness and acceptance within the medical community.

   Research to identify and validate surrogate endpoints is strongly recommended. Suitable surrogate endpoints would ideally be related to the biological processes underlying brain injury, and would predict future status. Desirable features in a surrogate would include (1) quantifiability, (2) correlation with degree of brain damage, (3) dynamic response with time, (4) relation to drug's mechanism of action, and (5) technical ease of use. A number of measures currently in use, such as intracranial pressure, therapy intensity level, (SiVO2), imaging, magnetic resonance spectroscopy, (DiH*C), (MTF), dialysis, and neuro-worsening/neuro-improvement, meet some of these criteria. No known surrogate is perfect, and none could serve as a primary endpoint in a Phase III trial.

   From the FDA's point of view, a surrogate may be used if it reasonably predicts clinical benefit. Ideally the surrogate does not just correlate with outcome, and treatment affects the prediction made by the surrogate. Several pitfalls attend reliance on surrogate outcomes: a treatment may affect the surrogate, but not the disease mechanism; a treatment may affect outcome in unintended negative ways, although it is positive on the surrogate; the surrogate is specific to a particular drug or class of compounds; there are time-dependent changes in the surrogate. Surrogates are useful if the endpoint is very difficult to study, the safety database is large, the disease mechanism is well understood, the surrogate occurs late in the pathophysiological pathway, and the surrogate has been studied in animal models. Research should be devoted to the development of biochemical/physiological markers. These markers should be related to damage and should dynamically reflect the course of injury and repair. Markers either in blood or CSF would be ideal, and possible candidates include S100, metabolites of lipid peroxidation, and tau proteins.

   Most clinical trials have focused on severe head injury; however, mild and moderate head injury can result in long-lasting effects, that range from headache to memory loss to the inability to return to work. Research priority should be given to developing a statistical tool for the analysis of mild and moderate head injury.

5. Confirm appropriate and timely drug delivery: Due to the blood brain barrier, it is necessary to show that drugs targeted to brain tissue are transported into the brain and reach sufficient concentrations to be active. Brain levels of drug should ideally be related to outcome. It is critical to know and appreciate the therapeutic window of a drug or intervention. In the clinical trials reviewed in the workshop, dosing with study drug occurred later after injury and continued for a longer length of time than in preclinical studies. Based on the animal studies, this mismatch would be expected to be both less effective and more toxic. The critical window of opportunity needs to be defined for each treatment, and the relationship between critical windows in rodents, larger animals, and humans needs to be established before launching efficacy trials.

6. Informed consent Obtaining informed consent is a key factor in head injury trials, since the patient may be unconscious or unable to give consent, a relative cannot be found, and time is of the essence. A waiver of consent may make it possible to enroll patients faster. Waivers require community consultation and public disclosure, requirements that are not easy to fulfill. Recommendations were made to streamline consent issues, and to form alliances with other groups affected by consent guidelines, such as epilepsy interest groups, critical care physicians, and the NIH bioethics committee.

Suggested guidelines for pharmaceutical sponsors

Several guarantees for scientists conducting trials with pharmaceutical industry support were considered to be important. These included: (1) independent and aggressive safety monitoring, (2) a commitment by the sponsor not to stop the trial early, but to go to completion, (3) an independent full analysis of the data, and (4) completion of analysis and publication of results within a reasonable length of time (1-2 yrs).

Two new trials proposed

1. A new fluid therapy trial. A phase II safety and feasibility trial was proposed to compare the standard of care against the standard of care plus hypertonic saline. This trial will test the hypothesis that treatment with hypertonic saline improves ICP, defined as fewer patients with ICP>30 for >30 min (or perhaps fewer patients with ICP >20 for a certain percentage of time?). Secondary endpoints would be the percentage of time that ICP >25, therapy intensity level, initial ICP, GOS, and neuro-worsening. An assessment of long-term cognitive function might also be warranted. Strict treatment sequence and patient management parameters will be provided. A waiver of informed consent will be requested for this trial, since it is anticipated that treatment ought to begin as soon after injury as possible (within 1-2 hours). Inclusion criteria would be closed head injury, age between 16-70 years, Glasgow Coma Score (GCS) of 3-12, and 1 reactive pupil. Hemodynamic instability secondary to systemic injury would exclude the patient. Patients would be stratified into groups based on GCS of 3-4, 5-8, and 9-12. Lessons that would be learned from this study include: (1) Can trauma centers be taught to manage fluids the same way? (2) Is hypertonic saline administered acutely and for 5 days safe? (3) Is hypertonic saline as good or better than mannitol?

2. An epidemiology study. It is estimated that 450,000-500,000 cases of traumatic brain injury occur each year. This translates to a prevalence of 330/100,000. From a survey of 4 states, the CDC has estimated that the prevalence is 220/100,000. Now that the census has been completed, and more exact incidence rates can be obtained from trauma centers in the American Brain Injury Consortium network, it is imperative to get good prevalence data for the year 2000.

¹Teasdale G and Jennett B Lancet, 1974, 2, 81. Five possible outcomes are described: death, vegetative, severe disability, moderate disability, good. For many trials the scale is "dichotomized" into unfavorable (death, vegetative, severe disability) and favorable (moderate disability or good) outcomes.

²Reference Guidelines for Treatment by AANS.

V. Participants

List of participants with affiliation.

Beth Ansell, Ph.D.
NIH/NCMRR
Bethesda, MD

Michael Badellino, M.D.
Temple University Hospital
Philadelphia, PA

Alex Baethmann
Ludwig-Maximilians University of Munich
Munich, Germany

Anat Biegon*, Ph.D.
Lawrence Berkeley National Laboratory
Berkeley, CA

Michael Bracken*, Ph.D.
Yale University School of Medicine
New Haven, CT

Ross Bullock*, M.D.
Virginia Commonwealth University
Richmond, VA

Mary Ellen Michel†, Ph.D.
NIH/NINDS
Bethesda, MD

Sung Choi*, Ph.D.
Virginia Commonwealth University
Richmond, VA

Guy Clifton*, M.D.
University of Texas HSC
Houston, TX

Charles Contant*, Ph.D.
Baylor College of Medicine
Houston, TX

William Coplin, Ph.D.
Wayne State University School of Medicine
Detroit, MI

Dalton Dietrich, Ph.D.
University of Miami School of Medicine
Miami, FL

Jamshid Ghajar*, M.D., Ph.D.
Brain Trauma Foundation
New York, NY

Sean Grady, M.D.
Hospital of the University of Pennsylvania
Philadelphia, PA

Robert Grossman, M.D.
Baylor College of Medicine
Houston, TX

Edward Hall*, Ph.D.
Parke-Davis Pharmaceutical Research
Ann Arbor, MI

William Heetderks, M.D., Ph.D.
NIH/NINDS
Bethesda, MD

Dave Hovda, Ph.D.
University of California Los Angeles
Los Angeles, CA

Jack Jallo, M.D.
Temple University School of Medicine
Philadelphia, PA

Russell Katz*, M.D.
Food and Drug Administration
Rockville, MD

Nachshon Knoller*, M.D.
Sheba Medical Center
Tel-Hashomer, ISRAEL

Patrick Kochanek, M.D.
University of Pittsburgh School of Medicine
Pittsburgh, PA

Andrew Maas*, M.D.
Academic Hospital
Rotterdam, The Netherlands

Jeannine Majde, Ph.D.
Office of Naval Research
Arlington, VA

Donald Marion, M.D.
University of Pittsburgh
Pittsburgh, PA

Tony Marmarou*, Ph.D.
Virginia Commonwealth University
Richmond, VA

Lawrence Marshall*, M.D.
University of California, San Diego
La Jolla, CA

Tracy McIntosh*, Ph.D.
University of Pennsylvania
Philadelphia, PA

Emmy Miller, RN, Ph.D.
University of Texas HSC
Houston, TX

Noel Mohberg, Ph.D.
Parke-Davis Pharmaceutical Research
Ann Arbor, MI

Paul Muizelaar*, M.D., Ph.D.
University of California at Davis Med. Center
Sacramento, CA

Raj Narayan†, M.D.
Temple University School of Medicine
Philadelphia, PA

Lawrence Pitts, M.D.
University of California at San Francisco
San Francisco, CA

Peter Quinn
Brain and Trauma Foundation
New York, NY

Gad Riesenfeld, Ph.D.
Pharmos Corporation
Iselin, NJ

Claudia Robertson*, M.D.
Baylor College of Medicine
Houston, TX

Kenneth Strauss, Ph.D.
Temple University School of Medicine
Philadelphia, PA

Graham Teasdale*, M.D.
Southern General Hospital
Sacramento, CA

Nancy Temkin*, Ph.D.
University of Washington
Seattle, WA

Ronald Tuma, Ph.D.
Philadelphia, PA

Charles Wade*, Ph.D.
Consultant
Sausalito, CA

Michael Walker, M.D.
NIH/NINDS
Bethesda, MD

Michael Weinrich, M.D.
NIH/NCMRR
Bethesda, MD

John Whyte*, M.D., Ph.D.
Temple University School of Medicine
Philadelphia, PA

Jack Wilberger, M.D.
Allegheny General Hospital
Pittsburgh, PA

Byron Young*, M.D.
University of KY Chandler Medical Center
Lexington, KY

Lorraine Yurkewicz*, Ph.D.
Pfizer Pharmaceuticals
Groton, CT

*speaker
†organizer